对SARS和COVID-19疫苗设计的线性B细胞表位预测:整合平衡的集体学习模型和重新采样策略
1Department of Management Information Systems, Faculty of Economics and Administrative Sciences, Karadeniz Technical University, Trabzon, Turkey.
PeerJ. Computer science
|June 26, 2025
概括
机器学习准确地预测B细胞表位和对SARS和COVID-19的抗体识别. 先进的重新采样和组合方法显著改善了预测性能,有助于疫苗开发.
科学领域:
- 免疫信息学是指免疫信息学.
- 机器学习 机器学习
- 计算生物学 计算生物学
背景情况:
- 准确预测B细胞表位和抗体识别对于理解SARS和COVID-19病原体至关重要.
- 数据不平衡在开发这些表征的强有力的预测模型方面构成了重大挑战.
研究的目的:
- 开发和验证一种机器学习框架,用于更好地预测B细胞表位和在SARS和COVID-19中的抗体识别.
- 用各种重新采样技术解决数据不平衡问题,并评估它们对模型性能的影响.
主要方法:
- 应用过量采样,不足采样和混合采样技术来重新平衡数据集.
- 使用集体分类器 (提升,包装,平衡) 与超参数优化 (GridSearchCV) 和特征选择 (RFE).
- 使用七个指标评估模型性能,包括ROC AUC,PR AUC和MCC,并进行统计学意义测试.
主要成果:
- 合成少数民族过量采样技术编辑的最近邻居 (SMOTE-ENN) 和ExtraTrees的组合实现了最高的ROC AUC0.9899.
- 使用ExtraTrees的实例硬度值 (IHT) 也表现出强的表现 (ROC AUC 0.9799).
- 统计分析证实了模型预测的显著改进.
结论:
- 将重新采样技术与平衡组合分类器集成,有效地提高了B细胞表位预测和SARS和COVID-19的抗体识别精度.
- 确定了潜在的表位候选人,为疫苗开发做出贡献,并推进免疫信息学研究.
更多相关视频
相关概念视频
Cross-reactivity
31.5K
Overview
31.5K
Conjugated Proteins
18.6K
Simple proteins and protein complexes contain only amino acids. In contrast, many other proteins, called conjugated proteins, covalently bond with non-protein moieties.
Nucleoproteins are protein complexes that contain nucleic acids, categorized as deoxyribonucleoproteins (DNPs) or ribonucleoproteins (RNPs) respectively. The nucleosome is a typical example of a DNP where nuclear DNA is associated with histone proteins. The major antigen for the Covid-19 virus SARS-CoV is an RNP that is critical...
Nucleoproteins are protein complexes that contain nucleic acids, categorized as deoxyribonucleoproteins (DNPs) or ribonucleoproteins (RNPs) respectively. The nucleosome is a typical example of a DNP where nuclear DNA is associated with histone proteins. The major antigen for the Covid-19 virus SARS-CoV is an RNP that is critical...
18.6K


